A set-membership algorithm based parameter identification method for lithium-ion batteries

The accuracy of lithium-ion battery model is one of the most important factors that affects the applicability of power battery in electrical vehicles. Based on the traditional forgetting factor recursive least square (FFRLS) method, the random noise should be subjected to the normal distribution of zero mean and zero covariance, which, however, is very difficult to be satisfied in practical application. In this paper, based on the first-order RC equivalent circuit model, the identification of lithium-ion battery model parameters is performed by using the set-membership identification algorithm with unknown but bounded noise. The model parameters are identified by the set-membership algorithm with the experimental data of UDDS test on the NCM battery module. Experiments and simulation results show that the new method can simulate the dynamics of battery well, it can keep terminal voltage error within 1%, alongside with the root mean square error(RMSE) improved up to 8% compared with the FFRLS, which verifies the feasibility and the effectiveness of the new method, as well as providing data support for accurate estimation of battery state.